Fig. 4

Application of 3DeFDR-5C to find cell type-specific looping interactions across three cellular states. a Heatmaps representing binned, matrix balanced 5C counts (Observed) around a known looping interaction between the Olig1 gene and an NPC-specific enhancer (chr16:91,135,612-91,330,612). Observed counts are divided by the computed local expected signal to obtain background-normalized counts (Observed/Expected). These counts are fitted with a logistic distribution and the resulting p-values are transformed into interaction scores, where interaction score = − 10*log2(p value). b Interaction scores are thresholded to isolate contacts that are differentially looping across cellular conditions and whose signal meets a baseline requirement for significance. This thresholding procedure is applied to both real and simulated null replicate sets to compute an eFDR estimate. The dynamic thresholding procedure is applied with increasing stringency until a user-specified target false discovery rate is reached. c Loop classifications obtained with 3DeFDR-5C in real (top) and simulated null (bottom) replicate sets shown in an interaction scatterplot representation. d, e Heatmap of final loop classifications at d individual bin-bin pairs and e classified looping clusters after applying 3DeFDR-5C at a threshold of 2%. f UpsetR scalable Venn diagrams for differential looping clusters called by 3DeFDR-5C at a target eFDR of 2%